yes, it is easily reproducible on local env by increasing the nova's config option metadata_cache_expiration value to something higher (i did 50 sec).
It fail in both place in verify_empty_devices()[1] as well as in verify_device_metadata()[2]
[1] https://github.com/openstack/tempest/blob/5aeb551a58078abd34884583ac70ad425f4590ba/tempest/api/compute/servers/test_device_tagging.py#L310
[2] https://github.com/openstack/tempest/blob/5aeb551a58078abd34884583ac70ad425f4590ba/tempest/api/compute/servers/test_device_tagging.py#L306
as matt mentioned in gerrit, we should go with wait loop here than hard coded sleep.
yes, it is easily reproducible on local env by increasing the nova's config option metadata_ cache_expiratio n value to something higher (i did 50 sec).
It fail in both place in verify_ empty_devices( )[1] as well as in verify_ device_ metadata( )[2]
[1] https:/ /github. com/openstack/ tempest/ blob/5aeb551a58 078abd34884583a c70ad425f4590ba /tempest/ api/compute/ servers/ test_device_ tagging. py#L310
[2] https:/ /github. com/openstack/ tempest/ blob/5aeb551a58 078abd34884583a c70ad425f4590ba /tempest/ api/compute/ servers/ test_device_ tagging. py#L306
as matt mentioned in gerrit, we should go with wait loop here than hard coded sleep.